API429 in Cursor: custom API key, base URL and model checks

Understand the scope of a custom key

A custom key is useful when you want to select a model available through API429 and account for its API usage separately. Begin with one short text question in local Chat. This avoids confusing connection setup with a large project task, tool permissions or a growing conversation history. Do not upload private code merely to see whether an endpoint accepts your key.

The reviewed Cursor documentation limits custom OpenAI keys to standard non-reasoning chat models. A published API429 model is therefore only a candidate for a local compatibility check. The example ID gpt-5.6-luna has a public chat entry, but that does not establish that its family lacks reasoning or that Cursor accepts it. If the editor refuses the ID, stop and check support for your specific build.

Prepare the endpoint and credential

Use https://gateway.api429.com/v1 as the custom OpenAI base URL. Enter your API429 client key in the editor credential field. Do not paste the full /v1/chat/completions URL into a field that expects a base address, and do not add a second /v1. Compare the exact model ID with GET /v1/models using your key before selecting it.

The public catalog describes published models, while the authenticated list determines access for your credential. The placeholder YOUR_API429_KEY is intentionally inert. Replace it only in the application credential field, and keep the real value out of screenshots, shared settings and support logs. Check the provider selection as well as the key if you see an authentication error.

Save the settings and run a bounded Chat check

Open Cursor Settings, then Models and the OpenAI API key section. Save your key and set the custom OpenAI base URL. Some builds label this control Override OpenAI Base URL; the reviewed documentation describes Save, while other versions may expose Verify. Follow the available interface in your installed build instead of importing an invented configuration file.

Add or enable gpt-5.6-luna only through the model controls that your version exposes. Select it in local Chat and ask a short nonsecret question. Confirm that you receive a response from the intended selection without an access error. Saving a key is not the same as a successful model call. Native Cursor execution was not performed for this guide; your check can incur API charges.

Keep Chat, Agent and built-in features distinct

A working text exchange does not prove tools, agent actions or every streaming parameter. Agent also depends on the request method and expected tool format. GPT 6.1 Sol, for example, cannot be promised just because an OpenAI-compatible endpoint exists: the reviewed Cursor restriction concerns non-reasoning models, and API429 documents that model tooling under Responses API.

Custom keys do not cover every Cursor feature. The reviewed documents exclude Tab, Auto, Cloud and Background Agents, Automations, Cursor CLI, API and SDK. Do not assume that an API429 key replaces the built-in models used by those modes. For a later agent experiment, use a disposable project and review proposed file and command actions before approval.

Troubleshoot safely and check separate charges

For 401, verify the credential and provider field. For 404, compare the base URL and actual request path. For an unknown model, compare both the key-specific catalog and Cursor model selection. If Chat succeeds while an Agent tool fails, identify the rejected feature before repeating a large request. Share only redacted error text, never the key or private project contents.

Cursor describes server-side prompt assembly for custom-key requests, so this setup does not mean all processing stays on your computer. Its documentation says the key is sent over an encrypted connection with each request and is not stored on Cursor servers. API429 usage and the Cursor plan remain separate costs. Review current plan terms, including the documented custom-key data-retention limits and any applicable Teams or Enterprise token charge.

Cursor field reference; this is not an importable settings file

Provider setting: OpenAI
API key: YOUR_API429_KEY
Custom OpenAI base URL: https://gateway.api429.com/v1
Model ID to check (not guaranteed by Cursor): gpt-5.6-luna

Payment and a $10 test

• Russian bank card. Top up your balance in the account.
• Invoice for companies. Bank transfer with closing documents.
• USDT and TON. Top up in crypto.

$10 test credit. Subscribe to our Telegram channel through the bot and get $10 on your balance for the first requests. Get $10 in the bot